AI Productivity and Bounced Checks – February 13, 2026

Dear Valued Clients and Friends,

In this week’s Dividend Cafe:

• We look at the headwinds AI investors face, or at least vulnerabilities in the investment thesis, too many have overlooked
• We analyze the need for real productivity to materialize out of AI, not just enhanced margins, not just enhanced activity, but meaningful, additive output
• And we question the prevailing wisdom that indiscriminately throwing money at it is the right solution.  We don’t question if that may work out for some in the end, but we do question whether the risk-reward trade-off of that approach has historically worked well and if it has been properly considered this time around.

Let’s jump into the Dividend Cafe …

Download Podcast Transcript

A Review of the AI Investment Story

I have been writing about AI and what it means from an investment standpoint for quite some time.  In our annual white paper projecting major themes for 2026, I made the case that this would be a year when vulnerabilities in AI and AI-adjacent investments would become more evident.  In November, I assessed how AI investing concerns could be prudently avoided.  In October, I looked at the historical and economic arguments for why AI would not create the level of permanent, structural job disruption many fear.  The Monday Dividend Cafe has long been full of updates about the Mag-7, the evolution of AI, and AI-adjacent investment action, and my fundamental arguments have stayed consistent for quite some time.

Today’s Dividend Cafe is not about the AI investment vulnerabilities I normally discuss.  But in a nutshell, my view continues to believe that:

1. Artificial Intelligence and the technologies driving it are real, and its transformative impact will prove to be substantial.  But …
2. The impact is way less known, quantifiable, predictable, specifically identifiable, and practically investable (at this time) than many seem to realize.
3. The major vulnerability is that the investable story is, thus far, entirely in “pick and shovel” companies making AI infrastructure, and not in anyone using it or monetizing it.
4. The economic model for those making chips to power it and those building the language models that companies and people use is circular, and effectively Ponzi-like (for now).
5. The major capital expenditures powering all of this lack economic rationalization any time soon, and perhaps ever, leaving hyper-scalers vulnerable to the consequences of malinvestment, and leaving those they buy computing power from vulnerable to the inevitability of declining orders.
6. The assumption that there will be a broad cultural and political embrace of the entire AI story should not be taken for granted.
7. The assumption that all the AI-related companies can win at once, versus creating an environment where some win, and some lose, is dangerous, unsupported by math, and contrary to the lessons of history.
8. The belief that China poses no competition to U.S. companies in the AI space is wishful, and while it may prove to be true, it does seem to indicate a low regard for the risks that may exist.
9. The financing embedded in the whole AI story, thus far, has largely been equity-driven – essentially, funded by the massive cash flows of the hyper-scalers, and that is inevitably going to change as (a) More than 100% of these companies cash flow has been committed in many cases, and (b) Many players in the space don’t have that luxury.  As the broad AI-space takes on debt funding to drive its growth, more capital is available for productive use, and more capital is at risk, changing the risk profile of the entire narrative.

I will stop at nine because a Top Ten list seems too cliche.

Now, beyond those investment ramifications of the current AI moment, let’s talk about something else that is not yet getting much discussion.

Is. It. Really. True. That. AI. Is. Making. Us. More. Productive ???

The Promise is Real

Many AI skeptics talk about the need for more electricity to power this whole thing, and question the likelihood of securing the power we need (affordably and within political realities) to make it happen.  I get the argument, and especially get the fact that there is and is going to continue to be more public resistance to data center builds, but I generally find scarcity arguments unconvincing, mostly because they always, always seem to be wrong.  The Malthusian view of the world is usually right until it is wrong, and generally speaking, when there is enough opportunity to do so, we find a way to produce, extract, and create the resources we need to make the opportunity happen.  I am not suggesting the path to power creation will be smooth, but I do not think the challenges of meeting power demand will prove to be the death knell of AI.

I am so skeptical about the way the AI investment thesis has rolled out to the investing public and so cognizant of the nine issues I raised in the preceding section that I completely understand people labeling me as an AI-skeptic or AI-bear.  Neither is true when one divorces those terms from the investment context.  I believe AI’s promise to increase efficiencies is real.  I don’t know enough to understand how this may be, but I do believe there is a real chance that AI increases the success rates of new drugs and medical advancements coming to market.  Even apart from AI’s role in improving the FDA approval system, I believe AI will reduce medical administration burden, and that means increasing medical productivity, and that means saving lives or enhancing one’s quality of life.  That promise means a great deal to me.  I don’t take it for granted.

Many have already seen ways in which AI has replaced cumbersome parts of their workflow.  There is no doubt that AI generative technology offers tremendous promise in a wide array of industries and sectors.  I am excited for much of this, even as I have unanswered questions about how a lot of it will work.

What is at Play

No one needs to deny the level of AI activity going on, and no one needs to deny the speed at which AI can do certain things relative to a pre-AI world.  But the question before us is not one of activity but output.  The difference between activity and output is a matter of value creation.  The question about the current AI moment is the value creation that comes from a genuine increase in productivity.  There are reasons to be skeptical that we have yet seen this promise materialize.

The volume of content AI can generate, and the speed with which it can generate it, does not mean more output.  We need to see valuable goods and services as the output from AI, and of course, as investors, we need to see more revenue, better profit margins, and better returns on capital.  If AI increases our volume of activity and the speed of that activity, but does not increase output, the productivity has not increased.  I imagine everyone reading this will say, “Oh come on, of course it will improve productivity eventually!”  I am pretty much in that camp, too, but I believe a sober assessment of the output, thus far, is needed.

Margins are not Output

If phone operators, customer service reps, proofreaders, and junior analysts are all replaced by AI functionality, it seems to me that businesses will improve profit margins as labor costs come down and a cheaper mechanism for certain functions materializes.  That is a different issue, though, than the new net value being created.  Higher margins do create a shift in investment opportunity (the company with newly improved margins now has a newly created investment return), but that is categorically different from real economic growth if it does not result in increased output.  Information processing is not value creation, in and of itself – it is a tool for value creation.  I believe higher profit margins are a good thing, especially when they enable more investment into tangible productivity.  But I do not believe this challenge or issue has been answered yet.

The Labs Chime In

This sobering report from MIT made waves a few months ago, suggesting that AI adoption in the workforce had dramatically increased but that the productivity increase from such increased adoption was nowhere to be found.  They see little structural change thus far and largely attribute the problem to a lack of customization and integration.  The report does a deep dive into the problem, but it more than anything is objectively measuring something – that 95% of companies’ use of AI is not, yet, enhancing productivity.  The report carefully analyzes how all of that can change, and my reading of it is optimistic that it will, but the results were stunning to those who assumed the massive investment in AI tools would already be bearing fruit.

The Stanford Social Media Lab believes a lot of the problem is due to something that has come to be known as “workslop” – when AI generation is doing a task, but with less substance than is needed.  AI is generating more work product, but creating a large number of situations where work has to be redone or re-evaluated for accuracy, depth, and enhanced conclusions.

This all matched the CFO survey late last year, where roughly 70% said they were not (yet) seeing a change in productivity because of AI.  They do, however, maintain an optimism that it will drive revenue growth once wider adoption takes place.

One can dig in all they want; there is no empirical evidence (yet) of AI-generated productivity enhancement.  There is evidence of an increase in profitability in select usage, and there is a compelling narrative for where this materialization of productivity may come, but it is a narrative thus far, not a fact in evidence.  People can say, “We need more time to adopt,” and I think that is perfectly fair.  But do market valuations reflect the fact that such broad adaptation might not be as easy as people think, and may not result in the nirvana they expect?

Past is Prologue

In late December/early January, I read the 1997 book, Burn Rate, by Michael Wolff (now known as a serial book writer about President Donald Trump, but in a past life, an online media start-up venturer who wrote a book about his failed attempt that was, candidly, fascinating to read).  This quote was not about AI, and it was not even about the 1999 and 2000 dotcom asininity of Pets.com and such.  It was about a much earlier phase of the mid-90’s tech revolution, and it stuck out to me for a number of reasons:

“Nobody knows what’s going on.  The technology people don’t know.  The content people don’t know.  The money people don’t know.  Whatever we agree on today will be disputed tomorrow.  Whoever is leading today, I can say with absolute certainty, will be adrift or transformed some number of months from now.  It’s a kind of anarchy.  A strangely level playing field.  The Wild West.”

~ Michael Wolff, Burn Rate, written in 1997 about the dotcom moment, over three years before the implosion

This was not a bullish comment, obviously.  But nor was it bearish.  It was descriptive of a general reality – not predictive of a particular outcome (good or bad).  It described something that proved to be very true – not that a bunch of companies would succeed or fail – but that there was absolutely no real clarity, plan, strategy, or specificity behind it all.  That can be okay.  That can work out for some companies and some investors.  But it is not what many people believed.  I remember that investing period like it was yesterday, and the risk-taking investment thesis (both for the entrepreneurs, start-ups, and companies themselves, as well as the investors putting money into the space) was one of two things:

1. Just buy anything and assume the momentum of the moment carries something higher with some dumber fool there on the other side to exit you at a higher price, OR
2. Believe a certain revenue model, or strategy, or acquisition, or growth plan, was attractive and going to play out (more fundamental and strategic investors).

You might think, given what you know of me, that I would have something negative to say about the first thesis and something positive to say about the second group.  But ironically, the only people who made money on that tech moment of the 1990’s were in the first group, or if people in the second group made money, they made it accidentally, as their theses and outlook never materialized as envisioned, but some things just went up anyway, or for different reasons. 

There are lots of differences between the tech moment of the 90’s and the AI moment of the mid-2020’s, but there are lots of similarities, too.  All I am suggesting here is that there is a first group in the AI moment that mirrors the first group described above from the 1990’s – just buy, close your eyes, and let the momentum and mania do the work.  It works where it works (timing is everything) and rips some people’s eyeballs out where it doesn’t (timing is everything).  But it is the second group I am more interested in, because I believe they think that they have a theory of the case that is in line with a theory of the case some of the companies involved have, and I am not sure that the companies have a theory of the case at all.  I think they have a very sincere and thoughtful high-level acknowledgement of the opportunity and a very real desire to be early, to be a leader, and not to be left behind.  But I do not think they know the business model that this all flows into for the future, and when an investor thinks they have a plan or strategy or vision about it all that goes deeper than any plan, strategy, or vision the actual actors have, I find that indicative of a very different risk-reward trade-off than most people understand.  I do not believe there is anything wrong with not knowing how wild, disruptive, and promising technology will play out.  I just believe investors should understand it for what it is – a vulnerable and wildly dynamic situation that does not offer the clarity of ROI or promise that we seem to think it does.

Michael Wolff, also from Burn Rate in 1997:

“You can’t say, Hey, what did you think was going on?  There’s a fire burning like crazy that we have to keep throwing dollar bills on.  And while that was true of this business and every other business in the new internet industry and while everybody knew it was true – that is, that cash was just being consumed at a rate and with an illogic that no one could explain, much less justify – you must never, never admit it.”

Daunting.

Conclusion

I look forward to, Lord-willing, seeing a sustainable increase in real GDP growth because of AI technology.  I want to see it.  If real wages begin to rise, if we see a real increase in output across all sectors of the economy, if we see purposeful human action enhanced by AI, then I am going to celebrate.

But electricity never created wealth by existing.  It created wealth when entrepreneurs produced from it – when they converted it into energy.  The judgment, wisdom, and stewardship required to increase productivity will not come from AI.  But when those things utilize AI as a tool in the process, not to replace prudent risk-taking, but as a tool, I believe good things will happen.  Understanding the difference between the chicken and the egg, the means and the end, the cause and effect, the primary and derivative – will be the pivotal understanding for investors in this AI moment.

Chart of the Week

The politics of the AI story are in very, very early innings

Quote of the Week

“Every bubble starts with a real trend”

~ Stanley Druckenmiller

* * *
Markets are closed on Monday for the President’s Day federal holiday.  We will be back with you on Tuesday with the Daily Recap of the Dividend Cafe.  I will be in New York City all weekend writing away on my new dividend growth book (coming this August), and feel reasonably optimistic that I will be ready to submit the final manuscript to my publisher next weekend.  I wish you all a wonderful three-day weekend, and welcome any questions or comments you have.  This is a tricky topic in a tricky time in history.  But it is not a boring one.  Navigating the excitement of it is one thing; navigating away from bad temptations is another.  To that end, we work.

With regards,

David L. Bahnsen
Chief Investment Officer, Managing Partner

The Bahnsen Group
thebahnsengroup.com

This week’s Dividend Cafe features research from S&P, Baird, Barclays, Goldman Sachs, and the IRN research platform of FactSet